Evaluation of POSSUM for Patients Undergoing Pancreatoduodenectomy

ABSTRACT

Background: The POSSUM score is a predictive scoring system for postoperative morbidity. Although numerous studies have validated its application in major abdominal surgery, few have exclusively considered pancreatic resections, which have unique complications that are costly and problematic. We examined whether POSSUM could accurately reflect the clinical outcomes in pancreatic resection. Methods: A total of 694 consecutive resections of the pancreatic head were performed between 1993 and 2010 at the Department of General, Thoracic, and Vascular Surgery at the University Hospital Dresden. The POSSUM score calculated for each case was compared with the observed morbidity. Relevance and predictive performance of the score were assessed; in particular, because of the poor calibration of the POSSUM predictions on the Dresden data, a new score was created that was externally validated on patient cohorts from two different centers for pancreatic surgery. Results: The goodness-of-fit analysis revealed that the POSSUM score was not well calibrated because the POSSUM-predicted morbidity rate was 58.9% on average whereas the observed morbidity rate was 43.4%. Discrepancies occurred particularly among the predicted high-risk patients, for whom the score actually overestimated the morbidity risk. Therefore, we adapted the score and complemented it with additional prognostic parameters. The new score was validated in a patient cohort from two other German centers and fitted better to the data. Conclusion: The new score, named PS-POSSUM (POSSUM in pancreatic surgery), fits the data better. However, the prediction ability remains rather poor. PS-POSSUM may still be helpful, as it draws attention to additional risk and protective factors in addition to those in the original POSSUM score.

Keywords: :

• risk score
• pancreatic surgery
• POSSUM score
• evidence-based medicine
• complications

Supplemental information 2. Parameters that were considered in the analysis.

Table

SI 1: Tumor classification from our tumor patient cohort.